Construction progress identification method and device based on multi-task joint optimization

Through the multi-task joint optimization method, the network prediction model is trained using construction site image data, combined with category and target positioning losses, the problem of low construction progress identification efficiency in the existing technology is solved, and more accurate construction progress prediction and model training is achieved. Effect.

CN118781440BActive Publication Date: 2024-12-03HANGZHOU NEWGRAND TECHNOLOGY CO LTD +2
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411267378.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-11
Publication Date
2024-12-03
Estimated Expiration
2044-09-11

AI Technical Summary

Technical Problem

The existing construction project progress identification technology is inefficient, especially in complex background information and multi-task environments, making it difficult to achieve accurate prediction of construction progress.

Method used

Using a multi-task joint optimization method, a network prediction model is constructed by acquiring and preprocessing the construction site image sample data, combining category losses and target positioning losses, the model is trained to improve the accuracy of construction progress prediction.

Benefits of technology

It improves the accuracy of construction progress prediction and the accuracy of model training, can more effectively identify and locate construction progress, reduce interference from complex backgrounds, and improves the ability to predict construction site progress.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118781440B_ABST
    Figure CN118781440B_ABST
Patent Text Reader

Abstract

The present invention discloses a construction progress recognition method and device based on multi-task joint optimization. The method includes: obtaining an image sample training data set; performing image preprocessing on each construction site image in the image sample training data set to obtain a construction site image to be processed; inputting each construction site image to be processed into a pre-built network prediction model to obtain a construction progress category prediction result output by the model; determining a category loss value according to the construction progress category prediction result and a true label value; and, determining a target positioning loss value according to the construction progress category prediction result; determining a multi-task fusion loss value according to the category loss value and the target positioning loss value; training the network prediction model according to the multi-task fusion loss value until the model convergence condition is met, and obtaining a construction progress prediction model for construction progress prediction. The above technical solution improves the accuracy of construction progress prediction.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of image data processing, and in particular to a construction progress recognition method and device based on multi-task joint optimization. Background Art

[0002] In the field of construction, it is usually necessary to determine the construction progress of construction projects in different time periods, so as to clarify the progress of the construction projects. The progress of construction projects is very important to the project entrusting party.

[0003] The existing progress of construction projects includes manual determination and automatic identification. Manual determination of construction progress is inefficient and requires a lot of human resources. The existing automatic identification technology is to perform image analysis on-site construction images, but due to differences in image acquisition equipment and image acquisition personnel, some images contain a lot of complex background information. Some construction site photos may contain background information that is not related to the construction site, such as hills and roads. In addition, the background of some construction site photos may also contain other construction sites at different progress, making it more difficult to accurately judge the construction progress of the current construction site. The existing construction site image analysis algorithm cannot solve the above problems well, resulting in the inability to accurately predict the progress of on-site construction. Summary of the invention

[0004] The present invention provides a construction progress identification method and device based on multi-task joint optimization to improve the prediction accuracy of the construction progress.

[0005] According to one aspect of the present invention, a construction progress identification method and device method based on multi-task joint optimization are provided, and the method comprises:

[0006] Obtain an image sample training data set; the image sample training data set includes at least one construction site image; the construction site image is annotated with a real label value of a real construction progress category;

[0007] Performing image preprocessing on each of the construction site images in the image sample training data set to obtain a construction site image to be processed;

[0008] Inputting each of the construction site images to be processed into a pre-built network prediction model to obtain a construction progress category prediction result output by the model;

[0009] Determining a category loss value according to the construction progress category prediction result and the true label value; and,

[0010] Determining a target positioning loss value according to the construction progress category prediction result;

[0011] Determining a multi-task fusion loss value according to the category loss value and the target positioning loss value;

[0012] The network prediction model is trained according to the multi-task fusion loss value until the model convergence condition is met, and a construction progress prediction model is obtained for construction progress prediction.

[0013] According to one aspect of the present invention, a construction progress prediction method is provided, the method comprising:

[0014] Acquire the construction site image to be predicted;

[0015] The construction site image to be predicted is input into a construction progress prediction model to obtain a construction progress category prediction result; wherein the construction progress prediction model is generated using the construction progress prediction method described in the above embodiment.

[0016] According to another aspect of the present invention, a construction progress identification device based on multi-task joint optimization is provided, the device comprising:

[0017] A training set acquisition module is used to acquire an image sample training data set; the image sample training data set includes at least one construction site image; the construction site image is annotated with a real label value of a real construction progress category;

[0018] An image preprocessing module is used to perform image preprocessing on each of the construction site images in the image sample training data set to obtain a construction site image to be processed;

[0019] A model prediction module is used to input each of the construction site images to be processed into a pre-built network prediction model to obtain a construction progress category prediction result output by the model;

[0020] a category loss value determination module, used to determine the category loss value according to the construction progress category prediction result and the true label value; and

[0021] A target positioning loss value determination module is used to determine the target positioning loss value according to the construction progress category prediction result;

[0022] A multi-task fusion loss value determination module, used to determine a multi-task fusion loss value according to the category loss value and the target positioning loss value;

[0023] The model training module is used to perform model training on the network prediction model according to the multi-task fusion loss value until the model convergence condition is met, thereby obtaining a construction progress prediction model for construction progress prediction.

[0024] According to another aspect of the present invention, a construction progress prediction device is provided, the device comprising:

[0025] A site image acquisition module, used to acquire the construction site image to be predicted;

[0026] The category prediction module is used to input the construction site image to be predicted into the construction progress prediction model to obtain the construction progress category prediction result; wherein, the construction progress prediction model is generated using the construction progress prediction method described in the above embodiment.

[0027] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0028] at least one processor; and

[0029] a memory communicatively connected to the at least one processor; wherein,

[0030] The memory stores a computer program that can be executed by the at least one processor. The computer program is executed by the at least one processor so that the at least one processor can perform the method described in any embodiment of the present invention.

[0031] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the method described in any embodiment of the present invention when executed.

[0032] The technical solution of the embodiment of the present invention inputs each construction site image to be processed into a pre-built network prediction model, determines the category loss value according to the construction progress category prediction result and the true label value, determines the target positioning loss value according to the construction progress category prediction result, and trains the network prediction model based on the category loss value and the task loss, thereby realizing the training of the construction progress prediction model. The category loss and task loss are taken into account in the model training process, so that the model output result is as consistent as possible with the true progress category, thereby improving the model training accuracy, and then improving the prediction accuracy of the construction progress.

[0033] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0035] Figure 1 is a flowchart of a construction progress identification method based on multi-task joint optimization provided according to Embodiment 1 of the present invention;

[0036] Figure 2 is a flowchart of a construction progress identification method based on multi-task joint optimization provided according to Embodiment 2 of the present invention;

[0037] Figure 3 is a flowchart of a construction progress prediction method provided according to Embodiment 3 of the present invention;

[0038] Figure 4 is a structural schematic diagram of a construction progress prediction model provided according to a fourth embodiment of the present invention;

[0039] Figure 5 is a structural schematic diagram of a construction progress identification device based on multi-task joint optimization provided according to Embodiment 5 of the present invention;

[0040] Figure 6 is a structural schematic diagram of a construction progress prediction device provided according to Embodiment 6 of the present invention;

[0041] Figure 7 It is a schematic diagram of the structure of an electronic device for implementing the method of an embodiment of the present invention. DETAILED DESCRIPTION

[0042] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0043] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0044] Embodiment 1

[0045] Figure 1 The flowchart of a construction progress identification method based on multi-task joint optimization provided in the first embodiment of the present invention is applicable to the case of predicting the progress of a construction site. The method can be executed by a construction progress identification device based on multi-task joint optimization. The construction progress identification device based on multi-task joint optimization can be implemented in the form of hardware and / or software. The construction progress prediction device can be configured in an electronic device. Figure 1 As shown, the method includes:

[0046] S110, obtaining an image sample training data set; the image sample training data set includes at least one construction site image; the construction site image is annotated with a real label value of a real construction progress category.

[0047] Among them, the construction site pictures can be obtained by collecting images of the on-site construction environment of different divisions and their corresponding sub-divisions through on-site image acquisition equipment. For example, the image acquisition equipment can be a mobile phone, a camera or a drone, etc. Specifically, it includes construction site photos at different angles, different lighting conditions and different environments. In the data collection process, cameras with different resolutions can be used to ensure the diversity of image quality and clarity.

[0048] The construction progress category may be the engineering category of the corresponding sub-divisional engineering. The categories of sub-divisional engineering may include foundation and base category, main structure category, building decoration and renovation category, building roof category, building water supply and drainage and heating category, building electrical category, ventilation and air conditioning category, etc.

[0049] Among them, the sub-division engineering categories included in the foundation and base category, that is, the construction progress category, can include earthwork engineering, foundation engineering, pile foundation engineering, and underground waterproofing, etc. The sub-division engineering categories included in the main structure category can include steel bar tying, formwork laying, concrete pouring, prefabricated structure construction, masonry structure, and steel structure, etc. The sub-division engineering categories included in the building decoration installation category can include floor engineering, plastering engineering, door and window engineering, ceiling engineering, decorative panels, curtain wall engineering, and painting engineering, etc. The sub-division engineering categories included in the building roof category can include roof leveling and roof thermal insulation and waterproofing, etc. The sub-division engineering categories included in the building electrical category can include building electrical. The sub-division engineering categories of the ventilation and air conditioning category can include ventilation and air conditioning.

[0050] For different sub-division engineering categories, that is, construction progress categories, multiple construction site images corresponding to the corresponding categories can be obtained, and image label values ​​can be annotated for the corresponding construction site images. Specifically, the construction progress category in the corresponding construction site image is used as the label data of the corresponding sample image.

[0051] S120 , performing image preprocessing on each construction site image in the image sample training data set to obtain a construction site image to be processed.

[0052] For example, the construction site images in the image sample training data set can be resized to convert the construction site images of different sizes into a preset size, for example, they can all be converted into 256*256 pixel images. In order to further improve the subsequent processing efficiency of the image sample data, the pixel values ​​of the construction site RGB (Red, Green, Blue) images can also be normalized to obtain the construction site images to be processed.

[0053] S130, inputting each construction site image to be processed into a pre-built network prediction model to obtain a construction progress category prediction result output by the model.

[0054] The network prediction model may be a CNN (Convolutional Neural Networks) model, or a VGG (Visual Geometry Group, a network model for image recognition or image generation) or a ResNet (Residual Network, a residual network model), etc.

[0055] Exemplarily, the construction site images to be processed in the image sample training set are input into a pre-built or preset network prediction model to obtain the prediction result of the construction progress category output by the model.

[0056] S140: Determine a category loss value according to the construction progress category prediction result and the true label value.

[0057] For example, the category loss value can be determined based on the cross loss function according to the construction progress category prediction result and the true label value. After updating the network parameters through back propagation, the prediction model can extract more critical visual features, so that the category corresponding to the maximum probability in the output probability of each construction progress category is as consistent as possible with the actual construction progress category manually labeled.

[0058] S150. Determine the target positioning loss value according to the construction progress category prediction result.

[0059] For example, according to the prediction results of the construction progress category, the target positioning loss value is determined based on the mean squared error (MSE) or log loss function. It can reflect the difference between the predicted results and the actual results during the model task training process, so as to continuously optimize the model parameters.

[0060] S160: Determine a multi-task fusion loss value according to the category loss value and the target positioning loss value.

[0061] Exemplarily, the loss and / or weighted loss sum of the loss value and the target positioning loss value may be determined as the multi-task fusion loss value.

[0062] S170. Perform model training on the network prediction model according to the multi-task fusion loss value until the model convergence condition is met, and obtain a construction progress prediction model for construction progress prediction.

[0063] Exemplarily, if the multi-task fusion loss value meets the preset model iteration completion condition, the model training is terminated to obtain a converged construction progress prediction model. The model iteration completion condition may be that the loss value tends to be stable or the loss value reaches a preset loss threshold, which is not limited in this embodiment. The construction progress prediction model is used to predict the construction progress of the construction site.

[0064] In a specific construction progress prediction scenario, obtain the construction site image to be predicted, input the construction site image to be predicted into the construction progress prediction model, and obtain the construction progress category prediction result. According to the construction progress category prediction result, determine at least one first-level construction progress category; according to each first-level construction progress category, determine the target construction progress category.

[0065] The construction site image to be predicted may be an image for which progress prediction of the construction site is to be performed, and may be specifically acquired by an on-site image acquisition device.

[0066] By inputting the construction site image to be predicted into the construction progress prediction model, the prediction result of the construction progress category output by the model can be obtained. Among them, the prediction result is the construction progress category. It should be noted that the construction progress category output by the model is the sub-division engineering category. The number of categories output by the model can be set in advance. For example, the model can output five construction progress categories with probability values ​​in the top five. Therefore, according to the sub-division engineering category output by the model, at least one first-level category of the construction progress, that is, the division engineering category, can be determined. For example, if the sub-division engineering category output by the model is the pile foundation engineering category, the corresponding first-level category is the foundation and foundation category.

[0067] If the model outputs five construction progress categories, the first-level construction progress categories corresponding to the five construction progress categories are determined respectively, and the one with the highest frequency is selected as the target construction progress category according to the five first-level construction progress categories. For example, if the five first-level construction progress categories are foundation and base category, foundation and base category, foundation and base category, main structure category, and foundation and base category, the target construction progress category can be determined to be foundation and base category.

[0068] The technical solution of the embodiment of the present invention inputs each construction site image to be processed into a pre-built network prediction model, determines the category loss value according to the construction progress category prediction result and the true label value, determines the target positioning loss value according to the construction progress category prediction result, and trains the network prediction model based on the category loss value and the task loss, thereby realizing the training of the construction progress prediction model. The category loss and task loss are taken into account in the model training process, so that the model output result is as consistent as possible with the true progress category, thereby improving the model training accuracy, and then improving the prediction accuracy of the construction progress.

[0069] Embodiment 2

[0070] Figure 2 A flowchart of a construction progress identification method based on multi-task joint optimization is provided in the second embodiment of the present invention. This embodiment is optimized and improved on the basis of the above technical solutions.

[0071] Furthermore, the network prediction model includes an image feature extraction network layer, a global average pooling layer and a linear classification layer; the step of "inputting each construction site image to be processed into a pre-built network prediction model to obtain a construction progress category prediction result output by the model" is refined into "inputting each construction site image to be processed into the image feature extraction network layer in the network prediction model for image feature extraction to obtain a first feature map prediction result; inputting the first feature map prediction result into the global average pooling layer in the network prediction model for feature dimensionality reduction to obtain a feature vector prediction result; inputting the feature vector prediction result into the linear classification layer in the network prediction model for image classification prediction to obtain a construction progress category prediction result." to improve the model structure of the network prediction model.

[0072] The step of "determining the target positioning loss value according to the construction progress category prediction result" is refined into "determining the second feature map prediction result according to the construction progress category prediction result and the first feature map prediction result; determining the first image pixel number of the foreground feature map and the first image pixel number of the background feature map according to the first feature map prediction result; and, determining the second image pixel number of the foreground feature map and the second image pixel number of the background feature map according to the second feature map prediction result; determining the first loss value based on the first feature map prediction result according to the second image pixel number of the foreground feature map and the second image pixel number of the background feature map; and, determining the second loss value based on the second feature map prediction result according to the first image pixel number of the foreground feature map and the first image pixel number of the background feature map; determining the task loss based on the first loss and the second loss." to improve the method of determining task loss.

[0073] It should be noted that for the parts not described in detail in the embodiments of the present invention, reference can be made to the descriptions of other embodiments. Figure 2 As shown, the method comprises the following specific steps:

[0074] S201, obtaining an image sample training data set; the image sample training data set includes at least one construction site image; the construction site image is annotated with a real label value of a real construction progress category.

[0075] S202: performing image preprocessing on each construction site image in the image sample training data set to obtain a construction site image to be processed.

[0076] S203, inputting each construction site image to be processed into the image feature extraction network layer in the pre-built network prediction model to extract image features, and obtaining a first feature map prediction result.

[0077] The image feature extraction network layer may be a convolutional neural network with residual connections, which is used to extract features from the image and obtain a feature response map. Specifically, the image feature extraction network layer may be generated based on ResNet-50, which is composed of a stack of multiple convolutional layers, and residual block connections are introduced between layers to avoid degradation problems during model training.

[0078] The feature response map obtained after the image feature extraction network layer performs image feature extraction on the construction site image to be processed is used as the first feature map prediction result.

[0079] S204: Input the first feature map prediction result into the global average pooling layer in the network prediction model to perform feature dimensionality reduction processing to obtain a feature vector prediction result.

[0080] Among them, the global average pooling layer is used for feature convolution processing. The pooling kernel size in the global average pooling layer is the same as the size of the input feature response map.

[0081] The global average pooling layer can be used to average the feature response map of each channel. For example, if the size of the input feature response map is M×N×D, where M and N represent the height and width of the feature response map, respectively, and D represents the number of channels, then the global average pooling layer can average the feature response map on each channel to obtain an output of size 1×1×D, which is the feature vector prediction result.

[0082] S205. Input the feature vector prediction result into the linear classification layer in the network prediction model for image classification prediction to obtain the construction progress category prediction result.

[0083] Among them, the linear classification layer is used to classify and predict the feature vector prediction results, and can be specifically composed of at least one fully connected layer. In this embodiment, 22 construction progress categories are specifically set, that is, 22 construction progress categories are predicted to obtain the prediction results of the construction progress categories, which can specifically include the probability values ​​corresponding to each construction progress category.

[0084] S206. Determine the category loss value according to the construction progress category prediction result and the true label value.

[0085] S207: Determine a second characteristic map prediction result according to the construction progress category prediction result and the first characteristic map prediction result.

[0086] Exemplarily, the fully connected layer weight value of each construction progress category can be obtained according to the construction progress category prediction result, and the fully connected layer weight value of the construction progress category corresponding to the feature response map of the first feature map prediction result is weight-multiplied with the feature response map to obtain the second feature map prediction result.

[0087] S208. Determine the number of first image pixels of the first foreground feature map and the number of first image pixels of the first background feature map according to the prediction result of the first feature map.

[0088] If the feature response map of the first feature map prediction result is set to , where u represents the u-th element in the feature response map in the prediction result of the first feature map. Then the first foreground feature map The determination method can be as follows:

[0089]

[0090] in, The prospect threshold is pre-set by relevant technicians based on actual experience or experimental values, for example, Can be set to 0.4.

[0091] First background feature map The determination method can be as follows:

[0092]

[0093] in, The prospect threshold is pre-set by relevant technicians based on actual experience or experimental values, for example, Can be set to 0.4.

[0094] The number of pixels of the first image of the first foreground feature map is , the number of pixels of the first image of the first background feature map is .

[0095] S209: Determine the number of second image pixels of the second foreground feature map and the number of second image pixels of the second background feature map according to the second feature map prediction result.

[0096] If the feature response map of the second feature map prediction result is set to , where u represents the u-th element in the feature response map in the prediction result of the second feature map. Then the second foreground feature map The determination method can be as follows:

[0097]

[0098] Second background feature map The determination method can be as follows:

[0099]

[0100] The number of pixels of the second image of the second foreground feature map is , the number of pixels of the second image of the second background feature map is .

[0101] S210, determining a first loss value based on the prediction result of the first feature map according to the number of second image pixels of the second foreground feature map and the number of second image pixels of the second background feature map.

[0102] Among them, the first loss value is determined as follows:

[0103]

[0104] in, represents a second foreground feature map; a number of pixels of the second image representing a second foreground feature map; Represents the prediction result of the first feature map, specifically a feature response map; represents a second background feature map; The number of pixels of the second image representing the second background feature map.

[0105] S211. Determine a second loss value based on a prediction result of the second feature map according to the number of first image pixels of the first foreground feature map and the number of first image pixels of the first background feature map.

[0106] Among them, the second loss value is determined as follows:

[0107]

[0108] in, represents a first foreground feature map; a number of pixels of the first image representing the first foreground feature map; Represents the prediction result of the second feature map, specifically a feature response map; represents a first background feature map; The number of pixels of the first image representing the first background feature map.

[0109] S212: Determine a target positioning loss value according to the first loss value and the second loss value.

[0110] Among them, the task loss The determination method can be as follows:

[0111]

[0112] in, represents the second loss value; Represents the first loss value.

[0113] S213: Determine a multi-task fusion loss value according to the category loss value and the target positioning loss value.

[0114] S214. The network prediction model is trained according to the multi-task fusion loss value until the model convergence condition is met, and a construction progress prediction model is obtained for construction progress prediction.

[0115] The technical solution of this embodiment further refines the model structure of the network prediction model by inputting each construction site image to be processed into the image feature extraction network layer in the network prediction model for image feature extraction, obtaining the first feature map prediction result, inputting the first feature map prediction result into the global average pooling layer in the network prediction model for feature dimension reduction processing, obtaining the feature vector prediction result, and inputting the feature vector prediction result into the linear classification layer in the network prediction model for image classification prediction, obtaining the construction progress category prediction result. By determining the classification loss and task loss, the joint optimization of the two tasks of the network prediction model is realized, the interference of the complex background is reduced, and the model can extract more robust visual features. By defining the foreground background threshold, the weak supervision target positioning task is helped to further accurately respond to the feature area, thereby helping the classification task to improve the accuracy; and the linear classification layer after improving the accuracy can output a more accurate feature response map, obtain a more accurate target positioning, thereby further improving the training accuracy of the network prediction model.

[0116] In order to further realize the visualization display of the construction progress area, a target construction area image of the construction site image can also be generated.

[0117] In an optional embodiment, the network prediction model also includes an attention response map prediction network layer; the attention response map prediction network layer is respectively connected to the linear classification layer and the image feature extraction network layer; accordingly, after the first feature map prediction result is input into the global average pooling layer in the network prediction model for feature dimensionality reduction processing to obtain the feature vector prediction result, it also includes: inputting the first feature map prediction result and the construction progress category prediction result into the attention response map prediction network layer, extracting the target weight parameter of the target category in the construction progress category prediction result by the attention response map prediction network layer, and performing weighted summation of the target weight parameter and the first feature map prediction result to obtain an attention response map; according to the attention response map, generating a target construction area image of the construction site image.

[0118] Exemplarily, the attention response map prediction network layer can be a CAM (Class Activation Map) network layer. Specifically, the target weight parameter value of the target category can be extracted from the construction progress category prediction result, wherein the target category can be the category with the largest predicted probability value of the predicted construction progress category. The target feature response map corresponding to the target category in the feature response map of the first feature map prediction result is weighted multiplied by the target weight parameter value to obtain an attention response map. The attention response map is superimposed on the corresponding construction site image, and the target construction area is extracted from the superimposed image according to a preset pixel threshold to obtain a target construction area image. Among them, the target construction area image contains a bounding box with a selected construction area, which is used to display the main construction site recognized by the image.

[0119] Embodiment 3

[0120] Figure 3 This is a flowchart of a construction progress prediction method provided in the third embodiment of the present invention. This embodiment can be applied to the situation of predicting the progress of a construction site. The method can be executed by a construction progress prediction device. The construction progress prediction device can be implemented in the form of hardware and / or software. The construction progress prediction device can be configured in an electronic device. Figure 3 As shown, the method includes:

[0121] S310, obtaining a construction site image to be predicted;

[0122] S320: Input the construction site image to be predicted into the construction progress prediction model to obtain the construction progress category prediction result.

[0123] The construction progress prediction model is generated by using the construction progress prediction method described in the above embodiment.

[0124] Specifically, an image sample training data set is obtained. The image sample training data set includes at least one construction site image; the construction site image is annotated with a real label value of the real construction progress category; image preprocessing is performed on each construction site image in the image sample training data set to obtain a construction site image to be processed. Each construction site image to be processed is input into a pre-built network prediction model to obtain a construction progress category prediction result output by the model.

[0125] The network prediction model includes an image feature extraction network layer, a global average pooling layer and a linear classification layer. Specifically, each construction site image to be processed is input into the image feature extraction network layer in the network prediction model for image feature extraction to obtain the first feature map prediction result; the first feature map prediction result is input into the global average pooling layer in the network prediction model for feature dimensionality reduction to obtain the feature vector prediction result; the feature vector prediction result is input into the linear classification layer in the network prediction model for image classification prediction to obtain the construction progress category prediction result.

[0126] According to the construction progress category prediction result and the true label value, the category loss value is determined. According to the construction progress category prediction result and the first feature map prediction result, the second feature map prediction result is determined; according to the first feature map prediction result, the first image pixel number of the first foreground feature map and the first image pixel number of the first background feature map are determined; and according to the second feature map prediction result, the second image pixel number of the second foreground feature map and the second image pixel number of the second background feature map are determined; according to the second image pixel number of the second foreground feature map and the second image pixel number of the second background feature map, based on the first feature map prediction result, the first loss value is determined; and according to the first image pixel number of the first foreground feature map and the first image pixel number of the first background feature map, based on the second feature map prediction result, the second loss value is determined; according to the first loss value and the second loss value, the target positioning loss value is determined. According to the category loss value and the target positioning loss value, the multi-task fusion loss value is determined; according to the multi-task fusion loss value, the network prediction model is trained until the model convergence condition is met to obtain the construction progress prediction model.

[0127] In a specific construction progress prediction scenario, obtain the construction site image to be predicted, input the construction site image to be predicted into the construction progress prediction model, and obtain the construction progress category prediction result. According to the construction progress category prediction result, determine at least one first-level construction progress category; according to each first-level construction progress category, determine the target construction progress category.

[0128] The construction site image to be predicted may be an image for which progress prediction of the construction site is to be performed, and may be specifically acquired by an on-site image acquisition device.

[0129] By inputting the construction site image to be predicted into the construction progress prediction model, the prediction result of the construction progress category output by the model can be obtained. Among them, the prediction result is the construction progress category. It should be noted that the construction progress category output by the model is the sub-division engineering category. The number of categories output by the model can be set in advance. For example, the model can output five construction progress categories with probability values ​​in the top five. Therefore, according to the sub-division engineering category output by the model, at least one first-level category of the construction progress, that is, the division engineering category, can be determined. For example, if the sub-division engineering category output by the model is the pile foundation engineering category, the corresponding first-level category is the foundation and foundation category.

[0130] If the model outputs five construction progress categories, the first-level construction progress categories corresponding to the five construction progress categories are determined respectively, and the one with the highest frequency is selected as the target construction progress category according to the five first-level construction progress categories. For example, if the five first-level construction progress categories are foundation and base category, foundation and base category, foundation and base category, main structure category, and foundation and base category, the target construction progress category can be determined to be foundation and base category.

[0131] The technical solution of this embodiment obtains the construction site image to be predicted, inputs the construction site image to be predicted into the construction progress prediction model, and obtains the construction progress category prediction result. Since the construction progress prediction model is obtained by inputting each construction site image to be processed into a pre-built network prediction model, the category loss value is determined according to the construction progress category prediction result and the true label value, and the target positioning loss value is determined according to the construction progress category prediction result. The network prediction model is trained based on the category loss value and the task loss, and the training of the construction progress prediction model is realized. The category loss and task loss are considered in the model training process, so that the model output result is as consistent as possible with the true progress category, thereby improving the accuracy of model training, and then improving the accuracy of construction progress prediction. Therefore, the construction progress category prediction result is determined based on the construction progress prediction model, and the accurate determination of the progress prediction result is realized.

[0132] Embodiment 4

[0133] This embodiment further provides a preferred embodiment based on the above embodiment. Figure 4 A structural schematic diagram of a construction progress prediction model is shown.

[0134] The construction site image is input into the image feature extraction network layer in the construction progress prediction model for feature extraction to obtain the first feature map prediction result. The first feature map prediction result is input into the global average pooling layer for feature dimension reduction processing to obtain the feature vector prediction result. The feature vector prediction result is input into the linear classification layer for image classification prediction to obtain the prediction result of the construction progress category. According to the construction progress category prediction result and the true label value, the category loss value is determined.

[0135] According to the prediction results of the construction progress category, the fully connected layer weight value of each construction progress category can be obtained, and the fully connected layer weight value of the construction progress category corresponding to the feature response map of the first feature map prediction result is weight-multiplied with the feature response map to obtain the second feature map prediction result. The attention response map prediction network layer extracts the target weight parameter value of the target category output by the linear classification layer, wherein the target category is the category with the largest predicted probability value of the predicted construction progress category. The target feature response map corresponding to the target category in the feature response map of the first feature map prediction result is weight-multiplied with the target weight parameter value to obtain the attention response map. The attention response map is superimposed on the corresponding construction site image, and the target construction area is extracted from the superimposed image according to the preset pixel threshold to obtain the target construction area image.

[0136] During the training process of the construction progress prediction model, the loss values ​​of two tasks are considered, which are recorded as classification loss and task loss respectively.

[0137] Specifically, according to the prediction result of the construction progress category and the true label value, based on the cross entropy loss function, the category loss value is determined. According to the prediction result of the construction progress category and the prediction result of the first feature map, the prediction result of the second feature map is determined; according to the prediction result of the first feature map, the number of first image pixels of the first foreground feature map and the number of first image pixels of the first background feature map are determined; and according to the prediction result of the second feature map, the number of second image pixels of the second foreground feature map and the number of second image pixels of the second background feature map are determined; according to the second image pixel number of the second foreground feature map and the second image pixel number of the second background feature map, based on the prediction result of the first feature map, the first loss value is determined; and according to the first image pixel number of the first foreground feature map and the number of first image pixels of the first background feature map, based on the prediction result of the second feature map, the second loss value is determined; and according to the first image pixel number of the first foreground feature map and the number of first image pixels of the first background feature map, based on the prediction result of the second feature map, the second loss value is determined; according to the first loss value and the second loss value, the target positioning loss value is determined.

[0138] If the feature response map of the first feature map prediction result is set to , where u represents the u-th element in the feature response map in the prediction result of the first feature map. Then the first foreground feature map The determination method can be as follows:

[0139]

[0140] in, The prospect threshold is pre-set by relevant technicians based on actual experience or experimental values, for example, Can be set to 0.4.

[0141] First background feature map The determination method can be as follows:

[0142]

[0143] in, The prospect threshold is pre-set by relevant technicians based on actual experience or experimental values, for example, Can be set to 0.4.

[0144] The number of pixels of the first image of the first foreground feature map is , the number of pixels of the first image of the first background feature map is .

[0145] If the feature response map of the second feature map prediction result is set to , where u represents the u-th element in the feature response map in the prediction result of the second feature map. Then the second foreground feature map The determination method can be as follows:

[0146]

[0147] Second background feature map The determination method can be as follows:

[0148]

[0149] The number of pixels of the second image of the second foreground feature map is , the number of pixels of the second image of the second background feature map is .

[0150] Among them, the first loss value is determined as follows:

[0151]

[0152] in, represents a second foreground feature map; a number of pixels of the second image representing a second foreground feature map; Represents the prediction result of the first feature map, specifically a feature response map; represents a second background feature map; The number of pixels of the second image representing the second background feature map.

[0153] Among them, the second loss value is determined as follows:

[0154]

[0155] in, represents the first foreground feature map; a number of pixels of the first image representing the first foreground feature map; Represents the prediction result of the second feature map, specifically a feature response map; represents a first background feature map; The number of pixels of the first image representing the first background feature map.

[0156] Mission Loss The determination method can be as follows:

[0157]

[0158] in, represents the second loss value; Represents the first loss value.

[0159] The performance of the construction progress prediction model in identifying construction progress under different construction environments can be comprehensively evaluated using indicators such as Accuracy and F1 Score (F1 score, an indicator used to evaluate the performance of classification models). These three indicators will be calculated on 22 construction progress categories and the first-level progress parent categories corresponding to 7 construction progress categories to evaluate the model's coarse-grained construction progress identification capabilities and fine-grained construction progress identification capabilities.

[0160] For the image sample training data set used for model training, 70% of them can be randomly selected as sample training sets and 30% as sample test sets. For all methods in the experiment, the Adam optimizer can be used for training with a batch size of 96 and the number of iterations can be set to 200.

[0161] In addition, a variety of data enhancement techniques are used, including random cropping, flipping, scaling, etc., to increase the diversity of training data and improve the generalization and robustness of the model. By introducing more sample changes, the model will better learn the characteristics of different construction environments, thereby improving the accuracy and robustness of model detection. The main feature of the construction progress prediction model of this embodiment is that it can locate the construction subject in the input image while outputting the classification prediction results, effectively avoiding the influence of complex background on the classification results.

[0162] Embodiment 5

[0163] Figure 5 The structure diagram of a construction progress identification device based on multi-task joint optimization provided in the fifth embodiment of the present invention. The construction progress identification device based on multi-task joint optimization provided in the embodiment of the present invention can be applied to the situation of predicting the progress of a construction site. The construction progress identification device can be implemented in the form of hardware and / or software, such as Figure 5 As shown, it specifically includes: a training set acquisition module 501, an image preprocessing module 502, a model prediction module 503, a category loss value determination module 504, a target positioning loss value determination module 505, a multi-task fusion loss value determination module 506 and a model training module 507. Among them,

[0164] The training set acquisition module 501 is used to acquire an image sample training data set; the image sample training data set includes at least one construction site image; the construction site image is annotated with a real label value of a real construction progress category;

[0165] An image preprocessing module 502 is used to perform image preprocessing on each of the construction site images in the image sample training data set to obtain a construction site image to be processed;

[0166] The model prediction module 503 is used to input each of the to-be-processed construction site images into a pre-built network prediction model to obtain a construction progress category prediction result output by the model;

[0167] A category loss value determination module 504 is used to determine a category loss value according to the construction progress category prediction result and the true label value; and

[0168] A target positioning loss value determination module 505 is used to determine a target positioning loss value according to the construction progress category prediction result;

[0169] A multi-task fusion loss value determination module 506, used to determine a multi-task fusion loss value according to the category loss value and the target positioning loss value;

[0170] The model training module 507 is used to perform model training on the network prediction model according to the multi-task fusion loss value until the model convergence condition is met, thereby obtaining a construction progress prediction model for construction progress prediction.

[0171] The technical solution of the embodiment of the present invention inputs each construction site image to be processed into a pre-built network prediction model, determines the category loss value according to the construction progress category prediction result and the true label value, determines the target positioning loss value according to the construction progress category prediction result, and trains the network prediction model based on the category loss value and the task loss, thereby realizing the training of the construction progress prediction model. The category loss and task loss are taken into account in the model training process, so that the model output result is as consistent as possible with the true progress category, thereby improving the model training accuracy, and then improving the prediction accuracy of the construction progress.

[0172] Optionally, the network prediction model includes an image feature extraction network layer, a global average pooling layer and a linear classification layer; accordingly, the model prediction module 503 includes:

[0173] A first feature map prediction unit is used to input each of the construction site images to be processed into the image feature extraction network layer in the network prediction model to extract image features and obtain a first feature map prediction result;

[0174] A feature vector prediction unit, used for inputting the first feature map prediction result into the global average pooling layer in the network prediction model to perform feature dimension reduction processing to obtain a feature vector prediction result;

[0175] The image classification prediction unit is used to input the feature vector prediction result into the linear classification layer in the network prediction model to perform image classification prediction and obtain the construction progress category prediction result.

[0176] Optionally, the target positioning loss value determination module 505 includes:

[0177] A second characteristic graph prediction unit, configured to determine a second characteristic graph prediction result according to the construction progress category prediction result and the first characteristic graph prediction result;

[0178] A first pixel number prediction unit is used to determine the number of first image pixels of a first foreground feature map and the number of first image pixels of a first background feature map according to the prediction result of the first feature map; and

[0179] A second pixel number prediction unit, used to determine the number of second image pixels of the second foreground feature map and the number of second image pixels of the second background feature map according to the prediction result of the second feature map;

[0180] a first loss value determining unit, configured to determine a first loss value based on a prediction result of the first feature map according to the number of second image pixels of the second foreground feature map and the number of second image pixels of the second background feature map; and

[0181] A second loss value determining unit, configured to determine a second loss value based on a prediction result of the second feature map according to the number of first image pixels of the first foreground feature map and the number of first image pixels of the first background feature map;

[0182] A target positioning loss value determining unit is used to determine a target positioning loss value according to the first loss value and the second loss value.

[0183] Optionally, the network prediction model further includes an attention response map prediction network layer; the attention response map prediction network layer is connected to the linear classification layer and the image feature extraction network layer respectively; accordingly, the model prediction module 503 further includes:

[0184] An attention response map generation module is used for inputting the first feature map prediction result into the global average pooling layer in the network prediction model to perform feature dimensionality reduction processing to obtain a feature vector prediction result, inputting the first feature map prediction result and the construction progress category prediction result into the attention response map prediction network layer, extracting the target weight parameter of the target category in the construction progress category prediction result by the attention response map prediction network layer, and performing weighted summation of the target weight parameter and the first feature map prediction result to obtain an attention response map;

[0185] The target construction image generation module is used to generate a target construction area image of the construction site image according to the attention response map.

[0186] The construction progress prediction device provided in the embodiment of the present invention can execute the construction progress prediction method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0187] Embodiment 6

[0188] Figure 6 The structure diagram of a construction progress prediction device provided in the fifth embodiment of the present invention is shown in FIG. A construction progress prediction device provided in the embodiment of the present invention can be applied to the situation of predicting the progress of a construction site. The construction progress prediction device can be implemented in the form of hardware and / or software, such as Figure 6 As shown, it specifically includes: a scene image acquisition module 601 and a category prediction module 602. Among them,

[0189] The on-site image acquisition module 601 is used to acquire the construction site image to be predicted;

[0190] The category prediction module 602 is used to input the construction site image to be predicted into the construction progress prediction model to obtain a construction progress category prediction result; wherein the construction progress prediction model is generated using the construction progress prediction method described in the above embodiment.

[0191] The technical solution of this embodiment obtains the construction site image to be predicted, inputs the construction site image to be predicted into the construction progress prediction model, and obtains the construction progress category prediction result. Since the construction progress prediction model is obtained by inputting each construction site image to be processed into a pre-built network prediction model, the category loss value is determined according to the construction progress category prediction result and the true label value, and the target positioning loss value is determined according to the construction progress category prediction result. The network prediction model is trained based on the category loss value and the task loss, and the training of the construction progress prediction model is realized. The category loss and task loss are considered in the model training process, so that the model output result is as consistent as possible with the true progress category, thereby improving the accuracy of model training, and then improving the accuracy of construction progress prediction. Therefore, the construction progress category prediction result is determined based on the construction progress prediction model, and the accurate determination of the progress prediction result is realized.

[0192] Optionally, the device further comprises:

[0193] A first-level progress category determination module, configured to determine at least one first-level construction progress category according to the construction progress category prediction result after inputting the construction site image to be predicted into the construction progress prediction model to obtain a construction progress category prediction result;

[0194] The target progress category determination module is used to determine the target construction progress category according to each of the first-level categories of the construction progress.

[0195] The construction progress prediction device provided in the embodiment of the present invention can execute the construction progress prediction method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0196] Embodiment 7

[0197] Figure 7A schematic diagram of an electronic device 70 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0198] like Figure 7 As shown, the electronic device 70 includes at least one processor 71, and a memory connected to the at least one processor 71 in communication, such as a read-only memory (ROM) 72, a random access memory (RAM) 73, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 71 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 72 or the computer program loaded from the storage unit 78 to the random access memory (RAM) 73. In the RAM 73, various programs and data required for the operation of the electronic device 70 can also be stored. The processor 71, the ROM 72, and the RAM 73 are connected to each other via a bus 74. An input / output (I / O) interface 75 is also connected to the bus 74.

[0199] A number of components in the electronic device 70 are connected to the I / O interface 75, including: an input unit 76, such as a keyboard, a mouse, etc.; an output unit 77, such as various types of displays, speakers, etc.; a storage unit 78, such as a disk, an optical disk, etc.; and a communication unit 79, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 79 allows the electronic device 70 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0200] The processor 71 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 71 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 71 executes the various methods and processes described above, such as the construction progress identification and / or prediction method.

[0201] In some embodiments, the construction progress identification and / or prediction method may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 78. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 70 via the ROM 72 and / or the communication unit 79. When the computer program is loaded into the RAM 73 and executed by the processor 71, one or more steps of the construction progress identification and / or prediction method described above may be performed. Alternatively, in other embodiments, the processor 71 may be configured to perform the construction progress identification and / or prediction method in any other appropriate manner (e.g., by means of firmware).

[0202] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0203] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0204] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, device, or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0205] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0206] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0207] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0208] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0209] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A construction progress identification method based on multi-task joint optimization, characterized in that: include: Acquire an image sample training data set; the image sample training data set includes at least one construction site image; The construction site image is annotated with a real label value of the real construction progress category; Performing image preprocessing on each of the construction site images in the image sample training data set to obtain a construction site image to be processed; Inputting each of the construction site images to be processed into the image feature extraction network layer in the network prediction model to extract image features, and obtaining a first feature map prediction result; Inputting the first feature map prediction result into the global average pooling layer in the network prediction model to perform feature dimensionality reduction processing to obtain a feature vector prediction result; Inputting the feature vector prediction result into the linear classification layer in the network prediction model to perform image classification prediction to obtain a construction progress category prediction result; Determining a category loss value according to the construction progress category prediction result and the true label value; and, Determining a second characteristic graph prediction result according to the construction progress category prediction result and the first characteristic graph prediction result; Determining the number of first image pixels of the first foreground feature map and the number of first image pixels of the first background feature map according to the prediction result of the first feature map; as well as, Determine the number of second image pixels of the second foreground feature map and the number of second image pixels of the second background feature map according to the prediction result of the second feature map; Determining a first loss value based on a prediction result of the first feature map according to the number of second image pixels of the second foreground feature map and the number of second image pixels of the second background feature map; as well as, Determining a second loss value based on a prediction result of the second feature map according to the number of first image pixels of the first foreground feature map and the number of first image pixels of the first background feature map; Determining a target positioning loss value according to the first loss value and the second loss value; Determining a multi-task fusion loss value according to the category loss value and the target positioning loss value; The network prediction model is trained according to the multi-task fusion loss value until the model convergence condition is met, and a construction progress prediction model is obtained for construction progress prediction.

2. The method according to claim 1, characterized in that The network prediction model further includes an attention response map prediction network layer; the attention response map prediction network layer is respectively connected to the linear classification layer and the image feature extraction network layer; accordingly, after the first feature map prediction result is input into the global average pooling layer in the network prediction model for feature dimensionality reduction processing to obtain the feature vector prediction result, it also includes: Inputting the first feature map prediction result and the construction progress category prediction result into the attention response map prediction network layer, extracting the target weight parameter of the target category in the construction progress category prediction result by the attention response map prediction network layer, and performing weighted summation of the target weight parameter and the first feature map prediction result to obtain an attention response map; A target construction area image of the construction site image is generated according to the attention response map.

3. A construction progress prediction method, characterized in that: include: Acquire the construction site image to be predicted; The construction site image to be predicted is input into a construction progress prediction model to obtain a construction progress category prediction result; wherein the construction progress prediction model is generated using the construction progress identification method based on multi-task joint optimization as described in any one of claims 1-2.

4. The method according to claim 3, characterized in that After inputting the construction site image to be predicted into the construction progress prediction model to obtain the construction progress category prediction result, the method further includes: Determine at least one first-level construction progress category according to the construction progress category prediction result; According to each of the first-level categories of construction progress, a target construction progress category is determined.

5. A construction progress identification device based on multi-task joint optimization, characterized in that: include: A training set acquisition module, used to acquire an image sample training data set; the image sample training data set includes at least one construction site image; The construction site image is annotated with a real label value of the real construction progress category; An image preprocessing module is used to perform image preprocessing on each of the construction site images in the image sample training data set to obtain a construction site image to be processed; A model prediction module is used to input each of the construction site images to be processed into a pre-built network prediction model to obtain a construction progress category prediction result output by the model; a category loss value determination module, used to determine the category loss value according to the construction progress category prediction result and the true label value; and A target positioning loss value determination module is used to determine the target positioning loss value according to the construction progress category prediction result; A multi-task fusion loss value determination module, used to determine a multi-task fusion loss value according to the category loss value and the target positioning loss value; A model training module, used for training the network prediction model according to the multi-task fusion loss value until the model convergence condition is met, thereby obtaining a construction progress prediction model for construction progress prediction; The network prediction model includes an image feature extraction network layer, a global average pooling layer and a linear classification layer; accordingly, the model prediction module includes: A first feature map prediction unit is used to input each of the construction site images to be processed into the image feature extraction network layer in the network prediction model to extract image features and obtain a first feature map prediction result; A feature vector prediction unit, used for inputting the first feature map prediction result into the global average pooling layer in the network prediction model to perform feature dimension reduction processing to obtain a feature vector prediction result; An image classification prediction unit, used for inputting the feature vector prediction result into the linear classification layer in the network prediction model to perform image classification prediction and obtain a construction progress category prediction result; The target positioning loss value determination module includes: A second characteristic graph prediction unit, configured to determine a second characteristic graph prediction result according to the construction progress category prediction result and the first characteristic graph prediction result; A first pixel number prediction unit is used to determine the number of first image pixels of a first foreground feature map and the number of first image pixels of a first background feature map according to the prediction result of the first feature map; and A second pixel number prediction unit, used to determine the number of second image pixels of the second foreground feature map and the number of second image pixels of the second background feature map according to the prediction result of the second feature map; a first loss value determining unit, configured to determine a first loss value based on a prediction result of the first feature map according to the number of second image pixels of the second foreground feature map and the number of second image pixels of the second background feature map; and A second loss value determining unit, configured to determine a second loss value based on a prediction result of the second feature map according to the number of first image pixels of the first foreground feature map and the number of first image pixels of the first background feature map; A target positioning loss value determining unit is used to determine a target positioning loss value according to the first loss value and the second loss value.

6. A construction progress prediction device, characterized in that: include: A site image acquisition module, used to acquire the construction site image to be predicted; A category prediction module is used to input the construction site image to be predicted into a construction progress prediction model to obtain a construction progress category prediction result; wherein the construction progress prediction model is generated using the construction progress identification method based on multi-task joint optimization as described in any one of claims 1-2.

7. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can perform the method according to any one of claims 1 to 2 and / or claims 3 to 4.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the method of any one of claims 1-2 and / or 3-4 when executed.